A Galaxy of Options

Astronomers can measure how much non-hydrogen glowing gas a galaxy has using a value called metallicity. This is a really useful metric in principle, Scudder says, because low metallicity values indicate unprocessed gas, and high metallicity values signify heavily recycled gas. Unfortunately, there isn’t a single, clean way of calculating that value. “We don’t have one method that works,” Scudder says. “We have a dozen methods that all kind of work in different contexts.” 

Scudder’s latest research, published in Monthly Notices of the Royal Astronomical Society with coauthors Aidan Khelil ’22 and Jordan Ordower ’25, aimed to make sure that metallicity values were still reliable coming from regions of a galaxy dominated by light from young stars, even as the data available to astronomers became massively more detailed.

In the past, astronomers only had access to a single spectra—a term that describes which wavelengths are present in light—per galaxy. Those wavelengths correspond to different elements, so it’s a way to identify what object (e.g., a star, a pulsar, a black hole) made the light. If this single spectra wasn’t dominated by starlight, astronomers would toss that galaxy and move on. 

Thanks to technology, galaxy images are more detailed; for example, the ones Scudder is working with have about 10,000 galaxies with anywhere between 200 and 2,000 spectra each. You can think of it like astronomers now seeing a 4K image instead of a blurry one. With this newly granular boundary, suddenly it mattered more to know if the dividing lines between starlight and not-starlight were affecting the metallicities.

Scudder took all the public spectral data for these galaxies, to the tune of 1.5 million data points, and ran them through 12 different methods for estimating metallicity values. “This is where data management becomes important, because you will ruin your computer if you ask it to plot 1.5 million things times 12,” she says. “If you ask it, the computer will go, ‘No,’ and shut down.” 

She then created a SQL database of these metallicity values and loaded them into computers in her lab for student researchers. Within these values, students looked at the image of each galaxy and where the boundary line changed, pixel by pixel, from “starlight” to “not starlight,” with each metallicity estimation method.

Scudder analyzed the boundary to find pixels with trusted metallicity values located right next to a pixel that didn’t seem like starlight. Since there should be a rotational symmetry in a galaxy, she rotated along that plane to find another pixel with the same metallicity value located the same distance from the center—but surrounded by trustworthy pixels.

“If there is not a difference, it tells me that being adjacent to something I don’t trust actually doesn’t matter at all,” she says. “This means the way we’ve been splitting these galaxies up is fine. But if I do see a difference, that tells me we have to be more careful about how we do our metallicity work going forward.” 

Her result was initially surprising. “What I found was that it matters some of the time,” Scudder says. For three of the 12 methods she tested, the boundary of star to not-star was getting contaminated, or displaying false positives. The common denominator? Those three methods calculated metallicity only using the ratio of nitrogen to hydrogen. 

Scudder suspected the problem was a sensitivity to nitrogen. Young stars have a “shell” of partially ionized material that includes a thin layer of nitrogen. But other celestial objects create different volumes of glowing nitrogen. For example, a supermassive black hole’s higher-energy light produces a thick shell of partially ionized nitrogen, which then permits a brighter nitrogen glow. 

The good news is that Scudder found that the metallicity methods that used elements in addition to nitrogen, like oxygen and sulphur, were not affected. Now that astronomers know this, they can either choose metallicities that aren’t so sensitive to nitrogen or just pay more attention to their boundary conditions. 

Scudder’s next project is figuring out how to compare these 12 metallicity methods to each other so that researchers can compare directly between the metallicities. Ultimately it means that the data and research up to this point is all OK. “It’s really good news for everything we have done so far,” Scudder says. “I found this result to be really reassuring in many ways.”